Sparse Activation Methods

Methods for inducing sparsity in feed-forward network activations to reduce inference costs in language models.

1 papers